development and testing of a miniaturized, dualfrequency ......development and testing of a...
TRANSCRIPT
Development and Testing of a Miniaturized, DualFrequency, SoftwareDefined GPS Receiver
for Space Applications
Andrew J. Joplin, E. Glenn Lightsey, Todd E. Humphreys
The University of Texas at Austin
February 1, 2012
02/01/2012
Andrew Joplin
2
Outline
● Motivation● Goals● Background● Initial Testing● OnOrbit Acquisition/Duty Cycling● LEO Navigation● Radio Occultation Observation● GEO Navigation● Hardware/Flight Testing● Conclusions
02/01/2012
Andrew Joplin
3
Motivation
● Why is there a need for a small, highprecision GPS receiver for space missions?
– Space science missions often require precise positioning– Use of legacy highprecision receivers on small satellites
restricted by volume, mass, and power requirements
● Why use small satellites for space science missions?– Low cost encourages university involvement– Large constellations of small satellites provide more instrument
coverage at a fraction of the cost
02/01/2012
Andrew Joplin
4
Goals
● <1 W OrbitAverage Power● <500 g Mass● 0.5U CubeSat Form Factor● SubMeter Low Earth Orbit (LEO) Navigation● Ionospheric Occultation Observation● Geosynchronous Orbit (GEO) Navigation
02/01/2012
Andrew Joplin
5
Background
02/01/2012
Andrew Joplin
6
Background: CASES
● CASES: Connected Autonomous Space Environment Sensor
– Softwaredefined, dualfrequencyreceiver
– Developed by the University ofTexas and Cornell University
– Designed to measure ionosphericscintillation
● Data Output– Navigation, observations, raw IQ, TEC, SV data
CASES: A Smart, Compact GPS Software Receiver forSpace Weather Monitoring. 2011 ION GNSS Conference
02/01/2012
Andrew Joplin
7
Background: FOTON
● FOTON: Fast, Orbital, TEC, Observables, and Navigation Receiver
– Spacebased, dualfrequency,softwaredefined receiver
– Developed from CASES– Hardware repackaged into
smaller form factor– Software altered to allow LEO
navigation
02/01/2012
Andrew Joplin
8
Background: FOTON
● Hardware (COTS components on custom boards)– Bobyn RF Front End– TI C6457 Digital Signal Processor
– Interface Board (ZBoard)
– Volume: 0.5U CubeSat form factor (8.3 x 9.6 x 3.8 cm)
– Mass: 326 g– Power: 4.5 W, <1 W orbit average power
● Software– Tracks GPS L1 C/A, L2C, and L5– Configurable for tracking other Lband signals
– Arbitrary number of channels, limited by data downlink
RF Front EndRF Front End
DSPDSP Z-BoardZ-Board
02/01/2012
Andrew Joplin
9
Background: Software Changes
● Terrestrial Spacebased Conversion:→● Release ITAR altitude/speed limits* – Done● Widen Doppler range to ±40 kHz (increases memory
requirements) – Done
● Radio Occultation:● Occultation prediction – In Process● Suppress clock fixup during occultation – Done● Openloop tracking – Done● Data bit prediction – Done
*Software uploads/testing done within ITARrestricted lab
02/01/2012
Andrew Joplin
10
Initial Testing
02/01/2012
Andrew Joplin
11
Initial Testing
● Testing on Spirent GPS signal simulator– Baseline receiver (Rx) testing– Ionosphere and Troposphere not simulated– Satellite (SV) clock and ephemeris errors not simulated
● Tests include:– Static simulation– Rectangular track (lowdynamics) simulation– Low earth orbit simulations
02/01/2012
Andrew Joplin
12
Initial TestingTerrestrial Tests
Static Simulation● 0.46 m RMS error
Rectangular Track Simulation● 0.83 m, 0.12 m/s RMS error
02/01/2012
Andrew Joplin
13
Tracked Doppler● Before software updates● Tracked 13 signals● ± 10 kHz Doppler range
Initial TestingLEO Doppler Test
Simulated Doppler● Inclined, 90 min. period LEO● Produced ± 40 kHz Doppler
02/01/2012
Andrew Joplin
14
Initial TestingLEO Benchmark Tests
● Simulated polar LEO● Doubledifference of observables
– Removes geometry and Rx clock effects
– Leaves only Rx and channelspecific noise
● RMS Errors:● Pseudorange: 0.1616 m● Carrier Phase: 0.5973 mm● Range Rate: 0.0569 m/s
02/01/2012
Andrew Joplin
15
OnOrbit Acquisition/Duty Cycling
02/01/2012
Andrew Joplin
16
OnOrbit Acquisition/Duty Cycling
● FOTON currently operates at 4.5 W
● <1 W orbitaverage power desired
● On/off duty cycling required
02/01/2012
Andrew Joplin
17
OnOrbit Acquisition/Duty Cycling
● Current onorbit acquisition capability:
– DSP reset time: 15 sec– FFTbased acquisition: <5 sec– Ephemeris retrieval: <30 sec– Overall time to first fix (TFF): <1 min.
● TFF dominated by DSP reset and ephemeris retrieval
– Store ephemerides in memory (in process)– Operate DSP in lowpower mode (in process)– TFF of a few seconds attainable
● Duty cycling is possible
02/01/2012
Andrew Joplin
18
LEO Navigation
02/01/2012
Andrew Joplin
19
LEO NavigationKalman Filter
● Extended Sequential Kalman Filter (EKF)● Combine L1 pseudorange and Doppler with
assumed LEO dynamics to smooth nav solution● State: ECEF position/velocity, Rx clock
bias/rate● State dynamics model:
● Pos/vel: J2 gravity model + noise● Clock: 1st order + noise
02/01/2012
Andrew Joplin
20
LEO NavigationKalman Filter
● Tested with LEO simulation● Comparison of EKF with pointwise linear
leastsquares solutions:
● Can be improved with higherfidelity dynamics model
Kalman Filter Point Solutions
RMS Position (m) 0.544 0.739
RMS Velocity (m/s) 0.0121 0.247
02/01/2012
Andrew Joplin
21
LEO NavigationDualFrequency Capability
● Ionospheric modelling algorithm:● Running estimate of TEC (Total Electron Content)
used to model ionosphere realtime● L2 pseudorange not otherwise used in nav solution
● LEO Test:● Polar LEO simulation● Ionosphere, L2C simulated
02/01/2012
Andrew Joplin
22
● Point Solution Results● RMS Errors:
– Pos: 1.47 m– Vel: 0.29 m/s
● Results can be improved with a Kalman filter that ingests L2C pseudorange
LEO NavigationDualFrequency Capability
02/01/2012
Andrew Joplin
23
Radio Occultation Observation
02/01/2012
Andrew Joplin
24
Radio Occultation Observation
● Rising/setting GPS satellite transmits through multiple layers of ionosphere
● GPS receiver on LEO satellite measures time history of ionospheric delay/total electron content (TEC)
http://www.cosmic.ucar.edu/
02/01/2012
Andrew Joplin
25
Radio Occultation Observation
● FOTON software already designed to measure TEC
● Dualfrequency LEO simulation demonstrates:
– Low elevation tracking– TEC estimation
● To do:
– Occultation prediction
Slant TEC vs. Elevation
02/01/2012
Andrew Joplin
26
GEO Navigation
02/01/2012
Andrew Joplin
27
GEO Navigation
● Geosynchronous Earth Orbit (GEO) outside of GPS orbit● Low signal strength, navigation very challenging
● FOTON GEO Simulation Results● Used OCXO + coherent accumulation● Unable to pull in side lobes● Tracked 24 SVs at a time over 2 hr period● RMS Errors:
– 10 m horizontal, 155 m vertical
– 0.75 m/s horizontal, 15 m/s vertical
● Better results attainable using data bit wipeoff● Already implemented, but not tested in GEO
http://www.gpsworld.com
02/01/2012
Andrew Joplin
28
Hardware/Flight Testing
02/01/2012
Andrew Joplin
29
Hardware/Flight Testing
● Completed:– Vibration testing– Thermal testing– Vacuum testing
● Upcoming:– Sounding rocket launch (Cornell): March 2012– Armadillo CubeSat launch (UT): 2014
02/01/2012
Andrew Joplin
30
Conclusions
● FOTON – a highprecision, adaptable, spacebased software receiver
● Duty cycling allows <1 W orbit average power● 326 g, 0.5U volume small enough for CubeSats● Kalman filter + dualfrequency meterlevel navigation→
● Low elevation tracking, TEC estimation demonstrates occultation observation potential
● Data bit wipeoff + long coherent integration GEO navigation →possible
● Upcoming test flights in 20122014
02/01/2012
Andrew Joplin
31
Acknowledgements
● UT Radionavigation Laboratoryradionavlab.ae.utexas.edu
● UT Center for Space Researchwww.csr.utexas.edu
● UT Satellite Design Laboratory● Cornell University
gps.ece.cornell.edu
02/01/2012
Andrew Joplin
32
Backup Slides
02/01/2012
Andrew Joplin
33
Motivation
● Why dualfrequency?– Increased precision using ionospherefree pseudorange– Direct computation of ionospheric delay
● Why softwaredefined?– Quick development – just recompile and test– Adaptable – use for navigation, ionospheric sensing, …– Reconfigurable onorbit
02/01/2012
Andrew Joplin
34
Kalman FilterBased POD
● LEO Simulation Testing● Repeat benchmark test simulation
Pos. Error (m) Vel. Error (m/s)
02/01/2012
Andrew Joplin
35
Kalman FilterBased POD
● RMS Errors:● Pos: 0.544 m
(vs. 0.739 m)● Vel: 0.0121 m/s
(vs. 0.247 m/s)● Can be improved
with more accurate dynamics model
Clock rate (m/s)
02/01/2012
Andrew Joplin
36
GEO Navigation
● High Earth Orbits (HEO) and Geosynchronous Earth Orbits (GEO) very challenging for GPS navigation
● Weak signals from GPStransmitter side lobes
● Very slow geometricchange
http://www.gpsworld.com
02/01/2012
Andrew Joplin
37
GEO Navigation
Description and Performance of the GPS Block I and II LBand Antenna and Link Budget. 1993 Institute of Navigation Conference, 1993.
● Solutions:● More stable clock (e.g. OCXO)
– Allows smaller PLL bandwidth (increases C/No)
● Long coherent integration of weak signals– Pulls in signal from GPS side lobe– Requires data bit wipeoff
● Kalman filtering
02/01/2012
Andrew Joplin
38
GEO Navigation
● Results